The rapid development of information technology, while promoting the wide dissemination of digital images and other multimedia information, has also given rise to a variety of image forgery means. The proliferation of forged images has put forward higher requirements for current image forgery techniques in terms of accuracy and fineness. This paper discusses the mathematical principle of CFAR target detection algorithm from two perspectives: the clutter statistical modeling of CFAR and CFAR detector, which is used as a target detection method for ordinary images. After completing the detection of ordinary images, the object shadow in ordinary images is taken as the entry point to explore the scene information provided by the object shadow. Based on the principle of planar homology, the co-point line constraint and intersection ratio consistency constraint methods are proposed to detect the image forgery region using the geometric features of the shadow. On the basis of the obtained image forgery regions, the image forgery model is constructed by integrating the traditional forgery methods such as ELA analysis method, spatial color method and CNN network algorithm. The designed forgery algorithm shows optimal performance in the forgery experiment of high-fake passports, and the leakage rate is as low as 0.288, which provides an effective technical reference for the identification of image forgery.
